Open Research Online
The Open University’s repository of research publications
and other research outputs
Ontology-driven document enrichment: principles,
tools and applications
Journal Item
How to cite:
Motta, Enrico; Buckingham Shum, Simon and Domingue, John (2000). Ontology-driven document enrichment:
principles, tools and applications. International Journal of Human-Computer Studies, 52(6) pp. 1071–1109.
For guidance on citations see FAQs.
c
2000 Academic Press
Version: Accepted Manuscript
Link(s) to article on publisher’s website:
http://dx.doi.org/doi:10.1006/ijhc.2000.0384
Copyright and Moral Rights for the articles on this site are retained by the individual authors and/or other copyright
owners. For more information on Open Research Online’s data policy on reuse of materials please consult the policies
page.
oro.open.ac.uk
To appear in the International Journal of Human-Computer Studies
Ontology-Driven Document Enrichment:
Principles, Tools and Applications
Enrico Motta, Simon Buckingham Shum and John Domingue
Knowledge Media Institute
The Open University
Walton Hall, MK7 6AA
Milton Keynes, UK
{e.motta, s.buckingham.shum, j.b.domingue}@open.ac.uk
Abstract. In this paper we present an approach to document enrichment, which
consists of developing and integrating formal knowledge models with archives of
documents, to provide intelligent knowledge retrieval and (possibly) additional
knowledge-intensive services, beyond what is currently available using 'standard'
information retrieval and search facilities. Our approach is ontology-driven, in the
sense that the construction of the knowledge model is carried out in a top-down
fashion, by populating a given ontology, rather than in a bottom-up fashion, by
annotating a particular document. In the paper we give an overview of the approach
and we examine the various types of issues (e.g., modelling, organizational and user
interface issues) which need to be tackled to effectively deploy our approach in the
workplace. In addition we also discuss a number of technologies we have developed
to support ontology-driven document enrichment and we illustrate our ideas in the
domains of electronic news publishing, scholarly discourse and medical guidelines.
1. INTRODUCTION
An important activity in knowledge management is "to convert text to knowledge" (O’Leary,
1998). This activity is central to knowledge management for two reasons: i) work practices and
information flow in organizations tend to be document-centred and ii) documents themselves do
not normally exhibit the amount of structure required to support semantically-aware search
engines or other forms of intelligent services. For these reasons there has been much interest in
technology to support the specification of structured information in textual documents, especially
web pages. The web standardisation community has focused on the underlying representational
infrastructure: XML (XML, 1999) has been proposed as the basic annotation formalism to
support the specification of structured information in web pages, while RDF builds on the XML
syntax to provide a standard declarative representation, which allows users to express semantic
relationships between items on the Web. Approaches such as Ontobroker (Fensel et al., 1998)
and Shoe (Heflin et al., 1998) provide formalisms and associated interpreters which make it
Ontology-Driven Document Enrichment. Page 2
possible to embed knowledge representation structures in web pages and use them to perform
inferences.
In this paper we look at the wider issues concerning "the conversion of text to knowledge" and
discuss a comprehensive approach to document enrichment (Sumner et al., 1998), which we are
trying out in a number of projects here at the Knowledge Media Institute. The approach is
characterized in terms of a set of activities, with associated informal guidelines. In the paper we
also describe a number of technologies, which we have developed to support our approach to
document-centred knowledge management. These technologies include a knowledge modelling
language
1
, form-based interfaces for adding and retrieving knowledge from a model, and a web-
based browser/editor, which supports the collaborative development of knowledge models over
the World-Wide-Web. Finally, we discuss the application of our approach to three domains:
electronic news publishing (Domingue and Motta, 1999), scholarly discourse (Buckingham
Shum et al., 1999) and medical guidelines (PatMan, 1998).
The paper is organized as follows: in the next section we give an overview of our approach, in
terms of the underlying methodological assumptions and the associated process model. In
section 3 we describe the technology we have developed to support the approach. In sections 4,
5 and 6 we discuss the application of the approach to the three aforementioned domains. Finally,
in sections 7 and 8 we discuss related work and reiterate the main contributions of this paper.
2. ONTOLOGY-DRIVEN DOCUMENT ENRICHMENT
Our approach is ontology-driven, in the sense that the construction of the knowledge model is
carried out in a top-down fashion, by populating a given ontology (Gruber, 1993), rather than in a
bottom-up fashion, by annotating a particular document. Figure 1 underlines this point
graphically, by emphasizing that the construction of a knowledge model is driven by a pre-
existing ontology, a set of documents and other sources of knowledge, such as appropriate
(human) experts. Following Gruber, we use the term “ontology” to indicate “a specification of a
reusable conceptualization”. More simply, an ontology can be seen as providing a vocabulary for
describing a range of models. For instance, an ontology for medical guidelines provides a
generic set of concepts and relations (e.g., medical condition, diagnostic guideline, guideline user
type), which can then be instantiated for particular guidelines to build guideline-specific models,
in domains such as stroke management or prevention of pressure ulcer.
An ontology-driven approach to model construction affords several advantages. Instantiating an
ontology is usually simpler and speedier than developing a model from scratch. In addition,
1
Here we use the term “knowledge modelling” as a short form for “knowledge-level modelling”, an expression
introduced by Allen Newell (1982) to describe models of knowledge-intensive behaviour which abstract from the
way this behaviour is implemented and focus instead on the knowledge employed by an agent and the goals the
agent is trying to achieve.
Ontology-Driven Document Enrichment. Page 3
because an ontology makes explicit the conceptualization underlying a particular model, it
becomes easier to maintain, reuse and interoperate the model with other components. Finally,
reasoning modules can be associated with an ontology and these are then applicable to all models
built by instantiating the ontology in question. For instance, in the case of medical guidelines,
one can envisage building ontology-specific guideline verification tools, which can then be used
to verify individual guidelines developed by instantiating the same generic guideline ontology.
Figure 1. Ontology-driven Document Enrichment
Because our model construction process is ontology-driven, we prefer to use the term
"enrichment" (Sumner et al., 1998), rather than "conversion" or "annotation", to refer to the
process of associating a formal model to a document (or set of documents). In general, a
representation, whether formal, graphical or textual, can be enriched in several different ways -
e.g., i) by providing information about the context in which it was created, ii) by linking it to
related artefacts of the same nature, or iii) by linking it to related artefacts of a different nature.
Although in our document-centred knowledge management work we provide multiple forms of
document enrichment, such as associating discussion spaces to documents (Sumner and
Ontology-Driven Document Enrichment. Page 4
Buckingham Shum, 1998), in this paper we will primarily concentrate on the association of
formal knowledge models to documents
2
.
Thus, an important facet of an ontology-centred approach to document enrichment is that the
formalised knowledge is not meant to be a translation of what is informally specified in the
associated document. Hence the knowledge model typically plays a different role from the
associated text. For instance, in the medical guideline scenario the knowledge model helps to
verify that all the kinds of knowledge expected to be found in a document describing a medical
guideline are indeed there. In the scholarly discourse scenario the knowledge model is meant to
capture the meta-knowledge required to structure academic debates (e.g., theory X contradicts
theory Y), which is often expressed only implicitly in publications (i.e., acquiring it typically
requires some interpretation effort) and is not modelled at all in traditional libraries. In a
nutshell, the emphasis in our approach is in identifying the added value (in terms of enabling
semantic retrieval and document indexing capabilities, or other reasoning services), which can be
provided by a formalised knowledge model. Our methodology comprises the following six steps.
1. Identify use scenario.
2. Characterize viewpoint for ontology.
3. Develop the ontology.
4. Perform ontology-driven model construction.
5. Customise query interface for semantic knowledge retrieval.
6. Develop additional reasoning services on top of knowledge model.
These steps are briefly described in the next sub-sections.
2.1 Identify Use Scenario
At this stage the services to be delivered by the knowledge management system are defined. In
particular, issues of feasibility and cost are investigated. Addressing the latter involves
answering questions such as: “What is the added value provided by the knowledge model,
considering the non-trivial costs associated with the development and instantiation of an
ontology?”, “Is there the need for a ‘full-blown’ knowledge model and for going beyond the
facilities provided by off-the-shelf search engines?”, “What additional reasoning services will be
provided, beyond deductive knowledge retrieval?”. Addressing feasibility issues requires
assessing (among other things) whether or not it is feasible to expect the target user community
to perform document enrichment or whether specialized human editors will be needed. This
latter solution introduces a significant bottleneck in the process and moreover assumes that to
introduce a central editor in the model development is actually feasible. This is definitely not the
case in some of our application domains. For instance, in the scholarly discourse scenario our
2
Having said so, the medical guideline scenario described in section 6 does integrate a formal model with a set of
discussion spaces, to provide multiple forms of document enrichment.
